Appeal No. 95-1217
Application 08/039,674
(e) eroding the complement image matrix with the
second structuring element to form a second
transformation matrix; and
(g) combining the corresponding points of the
first and second transformation matrices to form a
result matrix wherein each non-zero point identifies
an origin point where the shape has been recognized
in the first image matrix."
See col. 2, lines 15-31 (indentation added). Nothing about
this method suggests that it would be advantageous to use any
of these steps in a neural network system. Neural networks
operate in an entirely different manner than the Crimmins et
al. method. The
statement in Crimmins et al. which explains that prior art
machine vision systems had met with limited results does not
provide the motivation for doing image inversion in a neural
network system. Because the Examiner has not provided a
sufficient motivation to combine these references, we reverse
the rejection of claims 7 through 11.
c. Claims 12 through 20
The rejection of claims 12, and dependent claims 13
through 20, is based upon the Fukumizu reference teaching a
24
Page: Previous 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Next
Last modified: November 3, 2007